Probabilistic Latent Component Analysis for Radar Signal Detection

The detection of radar signal submerged in noise has always been substantial for radar performance. An algorithm of radar signal detection based on probabilistic latent component analysis is proposed in this paper. By employing probabilistic latent component analysis, signal spectrogram is explicitly modeled as a mixture of marginal distribution products and noise is described by a dictionary of marginals. The estimation of the most appropriate marginal distributions is performed using Expectation-Maximization algorithm. The goal of signal detection is achieved by selective reconstruction method of extracting signal from noise. Simulation results demonstrate the effectiveness of the proposed algorithm and the improvement of signal detection over wavelet detection.